Hi All,

Has anyone pursued a systematic method to writing kanji compounds? I was wondering if there were any decks that had prompts like:

card1:
Q: 野さい (vegetable)
A: 野菜 <and optional data>

card2:
Q: や球 (baseball)
A: 野球 <and optional data>

And so on. A minimal deck in which all Joyo Kanji are prompted only once is desirable (i.e. ~2100 card deck). The two-card example above really only prompts the user for "菜" and "野", but a total of four different kanji would be "processed". The prompt with would have the user write out the proper kanji for the reading that goes with the compound.

I imagine some Core6K hacking and spreadsheeting could produce a majority of this (and all the better that the vocab words be simple, familiar words), but my text parsing-fu is weak.

Any thoughts? Has this been discussed and one?

K.

I would advise against your focus on individual kanji. Individual kanji stop mattering once you've gotten a good handle on the basics. What starts to matter is real vocabulary. So don't bother with your 2100 deck idea. That would basically just be Heisig with Japanese keywords, and would be wasteful if you're already doing Heisig. If you want to switch to using Heisig with Japanese keywords then that would be good, but don't pretend this would give you something other than what Heisig is already giving you.

Here is how I've heard of people doing what you want to do, and how I plan to do it for myself.

1. Make a regular sentence/grammar/vocab deck. How heavily you advocate for sentences is up to you, but I would advise you to be testing example sentences for most of the vocabulary that exists in your deck. It will help with your reading comprehension. It will help with your understanding of grammar. It will give you a more natural understanding of the meaning of words. Good sources for sentences would be KO2001 or Kanji in Context. I use Kanji in Context which tends to have more complicated sentences.

2. Learn recognition of that sentence deck. So test being able to read, pronounce, and understand all words in all sentences in your deck.

3. Once a card gets firmly cemented in your brain(like not going to show up for another 6 months) then this is where you start testing production. You take a sentence, and split it so that it's testing for the production of a vocab words.

So this: 世の中には善人もいれば悪人もいると言うが、本当に悪意のある人はそんなに多くはないと思う。

Turns into: [よのなか]には[ぜんいん]もいれば[あくにん]もいると[いう]が、[ほんとう]に[あくい]のある[ひと]はそんなに[おおく]はないと[おもう]。

Now, at first you're gonna be really annoyed by this, but this study method has a lot of advantages. It's going to test your ability to write kanji, and you'll already be familiar with the meaning of the sentence because you've run it a bunch of times. The other thing it's going to test is okurigana like in 多く. Okurigana is tested for on lots of proficiency tests. The third thing is that if you have a sufficiently varied set of sentences then it will be a representative sample of general kanji usage. That is...you'll be testing writing the kanji that are used more frequently more often than you'll be testing writing kanji that are used less often.

Then when these production cards get to about 6 months then you can just have Anki randomly choose a version to give you when it pops up, and that's how you'll be able to maintain it long term. Anki has good support for doing cards with multiple sides and multiple ways. Experiment with it.

erlog Wrote:3. Once a card gets firmly cemented in your brain(like not going to show up for another 6 months) then this is where you start testing production. You take a sentence, and split it so that it's testing for the production of a vocab words.

So this: 世の中には善人もいれば悪人もいると言うが、本当に悪意のある人はそんなに多くはないと思う。

Turns into: [よのなか]には[ぜんいん]もいれば[あくにん]もいると[いう]が、[ほんとう]に[あくい]のある[ひと]はそんなに[おおく]はないと[おもう]。

Now, at first you're gonna be really annoyed by this, but this study method has a lot of advantages. It's going to test your ability to write kanji, and you'll already be familiar with the meaning of the sentence because you've run it a bunch of times. The other thing it's going to test is okurigana like in 多く. Okurigana is tested for on lots of proficiency tests. The third thing is that if you have a sufficiently varied set of sentences then it will be a representative sample of general kanji usage. That is...you'll be testing writing the kanji that are used more frequently more often than you'll be testing writing kanji that are used less often.

Then when these production cards get to about 6 months then you can just have Anki randomly choose a version to give you when it pops up, and that's how you'll be able to maintain it long term. Anki has good support for doing cards with multiple sides and multiple ways. Experiment with it.

Are there any fast way ways to make (or download) decks like these? I feel if I made them manually I'd be spending more time setting up the deck than actually studying. :/

If my computing ability were a bit better I'd try to set up a script to fetch example sentences and format them accordingly for each of the most common 25,000 common words or something.

erlog Wrote:So this: 世の中には善人もいれば悪人もいると言うが、本当に悪意のある人はそんなに多くはないと思う。

Turns into: [よのなか]には[ぜんいん]もいれば[あくにん]もいると[いう]が、[ほんとう]に[あくい]のある[ひと]はそんなに[おおく]はないと[おもう]。

Now, at first you're gonna be really annoyed by this, but this study method has a lot of advantages. It's going to test your ability to write kanji, and you'll already be familiar with the meaning of the sentence because you've run it a bunch of times.

I like this idea of using sentences you already have added...but in my opinion, this sentence is way too long for a card in an SRS. I agree that it shouldn't be 1 card = 1 kanji, but I think 1 card = 1 vocab word would be more doable. What you've got here are cards that are going to take a long time to get through, and you might just end up recognizing the sentence -- "oh it's this one. It's got 悪人, so that means あくい is 悪意！" rather than something shorter focusing on the word itself.

I used to have cards like this. This example uses edict definitions, because coupled with a (even short) sentence, it's enough to differentiate homophones
Q: あくい
本当に[あくい]のある人は少ない
ill will, spite, evil intention.....

A:
悪意

But then I came back to America, completely stopped handwriting Japanese, and the reviews started to annoy me, so I deleted them all. They were more time consuming than recognition cards, but it did help me remember how to write them.

Sorry, I should have mentioned that I'm pretty comfortable with reading and this isn't a post-RtK thing -- I finished RtK long ago, have finished Core6K just recently and am pretty comfortable with Kanji overall... It's just that if I think of all the words I have no problem at all recognizing and then ask myself "write the kanji for やさい (vegetable)" I get stumped. I get stumped on really really simple kanji. Now it's getting to the point that I have a visual memory of the shape of kanji involved in words and all that, but I draw a fat blank most of the time. Hence my motivation for writing compounds -- which I think is slightly different than keyword prompt-->kanji testing. The keyword in this case is the compound itself, so it's more like... ensuring that I know how to a majority of common kanji in at least one compound. Maybe it's fruitless, I dunno, but I'd rather have that ability or knowledge than being left with only remembering the Heisig keyword.

Asriel, your deck description is very much so what I'm looking for, I just wanted to make it a little bit easier by only ever having one kanji being tested at a time. I'm pretty sure if you were testing that one kanji you'd learn to pickup the associated kanji of the compound, especially if you're familiar with the word and its application/use.

K.

I should clarify -- this IS post RtK in that I've finished RtK already, but is not a "How do I start using kanji in real life now that I've finished RtK but don't know any Japanese grammar, words or blah blah blah" kind of thing

Asriel Wrote:I used to have cards like this. This example uses edict definitions, because coupled with a (even short) sentence, it's enough to differentiate homophones
Q: あくい
本当に[あくい]のある人は少ない
ill will, spite, evil intention.....

A:
悪意

I've recently started to do this, but I don't put the English definition. I try to grab a self explanatory sentence as best as I can, though, in order to differentiate homophones.

Edit
The reason I would suggest that you not do it one kanji at a time is that you will still have a hard time combining the word later on. You may find it difficult to actually grab the right order of characters when it comes time to use it. This is my thinking on it at least.

dat5h Wrote:I've recently started to do this, but I don't put the English definition. I try to grab a self explanatory sentence as best as I can, though, in order to differentiate homophones.

Yeah, mine really varied card-per-card. I used edict in this example just because I didn't want to open up anki to find a real card, so I just grabbed the Rikaichan definition.

Quote:You may find it difficult to actually grab the right order of characters when it comes time to use it.

Looks like you beat me to it. I tried that kind of card for a little while, but I soon noticed that a) there were generally 2 cards per vocab word, and b) I was screwing up the order of them all the time! Putting it all in one forces you to get both characters right, and in the correct order.


But I haven't done production cards for months now...I'm just looking back on experience.

Well writing the whole compound isn't something I'm opposed to (i.e. pure kana prompt within a sentence). As I'm done with Core6K it would be a good source. with all the available spreadsheet data it shouldn't be too hard to setup -- I would simply like to trim it down to a minimal set of cards covering all unique kanji in Core6K -- at least that way I'll be able to say the equivalent of "野菜の野" such phrases when trying to discuss kanji with natives and non-natives.

When you try to say "yo man, that's 'plains' because it's like.. computers... beforehand... you worked in the plains yo! and then it's 'vegetable', like a vulture in a TREE eating FLOWERS! you know what knaji I mean man!? That shit goes together and forms VEGETABLE in japanese, yasai!"

People are just like wtf dude.

K.

Asriel Wrote:

erlog Wrote:So this: 世の中には善人もいれば悪人もいると言うが、本当に悪意のある人はそんなに多くはないと思う。

Turns into: [よのなか]には[ぜんいん]もいれば[あくにん]もいると[いう]が、[ほんとう]に[あくい]のある[ひと]はそんなに[おおく]はないと[おもう]。

Now, at first you're gonna be really annoyed by this, but this study method has a lot of advantages. It's going to test your ability to write kanji, and you'll already be familiar with the meaning of the sentence because you've run it a bunch of times.

I like this idea of using sentences you already have added...but in my opinion, this sentence is way too long for a card in an SRS. I agree that it shouldn't be 1 card = 1 kanji, but I think 1 card = 1 vocab word would be more doable. What you've got here are cards that are going to take a long time to get through, and you might just end up recognizing the sentence -- "oh it's this one. It's got 悪人, so that means あくい is 悪意！" rather than something shorter focusing on the word itself.

I used to have cards like this. This example uses edict definitions, because coupled with a (even short) sentence, it's enough to differentiate homophones
Q: あくい
本当に[あくい]のある人は少ない
ill will, spite, evil intention.....

A:
悪意

But then I came back to America, completely stopped handwriting Japanese, and the reviews started to annoy me, so I deleted them all. They were more time consuming than recognition cards, but it did help me remember how to write them.

Your way is fine too, but I'm not going to do double the work by finding or making shorter versions of the sentences I already have. The sentences I'm working with are from Kanji in Context, and so if it's good enough for KiC then it's good enough for me. As well, that's not even long in comparison to some of the other sentences in my deck.

Example:
集団主義の日本では個人の能力よりもチームワークが強調される。そのため、学校教育でも子供たちはできるだけ個性を抑えて、みんなで協力して物事を進めるように教育される。これに対し、アメリカでは、一人一人の個性と能力に合った個別的な指導が重視される。

Long sentences in SRS are just fine. The majority of the cards I have are short sentences like the one you posted, but about 10% are long cards like this. It helps build reading and scanning speed, and on average it takes me about 40 minutes to run through 200 cards while checking understanding and pronunciation.